BUG: MultiIndex assignment fails to broadcast omitted levels

[bashtage]

• I have checked that this issue has not already been reported.
• (optional) I have confirmed this bug exists on the master branch of pandas.

Note: Not in a released pandas, only master

---

Code Sample

import pandas as pd
mi = pd.MultiIndex.from_product([["a","b"],[1,2],["z","y"]])
df = pd.DataFrame(index=mi,dtype=float)
mi2 = pd.MultiIndex.from_product([["a","b"],[1,2]])
s = pd.Series(index=mi2, dtype=float)
s.iloc[:]=3.14
df["new"] = s
print(df)

Problem description

In 1.3.0 MultiIndex assignment broadcast missing levels. As of #42231, this no longer works.

Expected Output

In master, I see

       new
a 1 z  NaN
    y  NaN
  2 z  NaN
    y  NaN
b 1 z  NaN
    y  NaN
  2 z  NaN
    y  NaN

In 1.3.0 or 1.2.5, I see

        new
a 1 z  3.14
    y  3.14
  2 z  3.14
    y  3.14
b 1 z  3.14
    y  3.14
  2 z  3.14
    y  3.14

Output of pd.show_versions()

``` INSTALLED VERSIONS ------------------ commit : f3a6753 python : 3.8.10.final.0 python-bits : 64 OS : Windows OS-release : 10 Version : 10.0.19043 machine : AMD64 processor : AMD64 Family 23 Model 113 Stepping 0, AuthenticAMD byteorder : little LC_ALL : None LANG : None LOCALE : English_United States.1252

pandas : 1.4.0.dev0+246.gf3a6753d2d
numpy : 1.20.2
pytz : 2021.1
dateutil : 2.8.1
pip : 21.1.3
setuptools : 52.0.0.post20210125
Cython : 0.29.23
pytest : 6.2.4
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : 7.22.0
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyxlsb : None
s3fs : None
scipy : 1.7.0
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : None

</details>

[trevorkask]

take

[trevorkask]

I have found two probable causes of this bug. However, I am not sure which one to go with because I understand that a patch led to this bug. So one option may be better in terms of not undoing the patch.

1. In the method _should_compare, line 5646, the parameters for the if statement check for boolean and numeric data, whereas the example given https://github.com/pandas-dev/pandas/issues/42557#issue-945839860, uses mixed data types(char and numeric). I suggest passing the is_mixed function rather than is_boolean or is_numeric.
2. In the method get_indexer, line 3481, I could change the 'and' to 'or', so that the should_compare condition gets bypassed. This is a bit riskier.

Kindly advise on the best way forward, depending on what the patch altered.

[bashtage]

I think the source of this bug was reverted in 4c9ef1b

[simonjayhawkins]

I think the source of this bug was reverted in 4c9ef1b

#42575